Founding AI Engineer

Ignizia; San Francisco, Northern (CA, KY); Hybrid; USD 180,000 - 260,000; Full time

Benefits offered by this job

Founding equity

Salary discussed early

Job summary

Ignizia is building the operating layer for human-AI-machine coordination and is hiring to own the agentic layer of the platform. You will craft the cross-platform AI companion, design extraction pipelines, and create agent workflows with human review gates to ensure honesty and reliability.

The role is hybrid in the San Francisco Bay Area, with in-person by default and travel to the design partner possible but not required.

Qualifications

• Senior AI/LLM engineer who has shipped agentic features to production: model orchestration, tool use, structured extraction and evals as a habit.
• Exceptional new grad with a research-to-product profile: translate research into usable, real-world tooling.
• Strong LLM application engineering: prompting, orchestration, retrieval and evaluation.
• Structured extraction experience: converting messy human input into typed, validated data.
• Production experience with Python and/or TypeScript in real projects.

Responsibilities

• Own the agentic layer of the platform and ensure it works end-to-end.
• Build the platform assistant and cross-platform AI companion.
• Develop extraction pipelines turning voice/conversation into structured data with provenance.
• Implement agent workflows with human review gates and truth-check mechanisms.

Skills

Python

TypeScript

LLM engineering

Structured extraction

Eval processes

Trust-focused product sense

Job description

Build the agentic layer of the platform, with honesty as a product feature.

San Francisco Bay Area Hybrid, in-person by default Senior or exceptional new grad Meaningful founding equity + salary

Agent workflows Structured extraction Evals

The mission

Make collaboration work between people, machines, and AI.

Twenty years ago organizations coordinated people. Today they coordinate people and software. Tomorrow they will coordinate people, AI agents, machines, and humanoids. Nothing in the modern stack was built for that. Organizations do not fail because people lack intelligence; they fail because that intelligence is not coordinated, and coordination failures already cost the economy roughly $2 trillion a year.

Ignizia is building the operating layer for human-AI-machine coordination. This is the unsolved layer of the AI transition: independent research keeps finding that capable AI agents lose much of their capability the moment they have to work together, and that the hardest problems in enterprise AI are organizational, not technical. Coordination cannot be downloaded. It has to be captured and shaped from how a real organization actually works, and whoever holds that teaming-context layer holds something no one else can import.

So we start at ground zero: small and mid-size manufacturing, the most coordination-dependent, least digitized work there is. We are live with a design partner, a 60-person luxury footwear manufacturer, where the platform is being shaped view by view against real operations. We are founder-led, early stage, and hiring our founding team.

The role

This is a separate discipline from our full-stack role, and it is where the company thesis gets proven or broken. The research is blunt: capable agents lose much of their capability the moment they have to work together. Your job is to build agents that do not, because they run on the one thing other agents lack: real teaming context, captured and shaped from a live organization. You own the agentic layer of the platform:

• The platform assistant: a cross-platform AI companion that helps workers, leads, and founders act on the living twin.
• Extraction pipelines: turning conversational and voice capture from real factory workers into structured, provenance-tagged organizational data, with measured quality.
• Agent workflows with human review gates: agents that draft, humans that confirm; nothing pretends to be more certain than it is.
• Honesty as a product feature: confidence scoring, fidelity levels, drafted-versus-verified lifecycles, and eval harnesses that measure whether the system tells the truth about what it knows.

What we are looking for

Two profiles genuinely fit this seat:

• A senior AI/LLM engineer who has shipped agentic features to production: model orchestration, tool use, structured extraction, and evals as a habit rather than an afterthought.
• An exceptional new grad with a research-to-product profile: you have taken research-grade work and turned it into something people use. For this profile the role opens once our senior full-stack anchor is in place, and an intern-to-hire path is available. We never hire a junior into an unanchored role; that is a promise about the mentorship you would get, not a hurdle.
• Strong LLM application engineering: prompting, tool design, orchestration, retrieval, and evaluation.
• Structured extraction experience: from messy human input to typed, validated data.
• Python and/or TypeScript in production.
• Product sense about trust: you care whether users can tell what the system actually knows.

How we will interview

We will give you a realistic capture transcript and the target schema and talk through how you would extract, score confidence, evaluate quality, and decide what a human must confirm. Bring opinions about evals.

What we offer

• A problem worth a decade: the coordination layer of the AI transition, started where it is hardest and most real.
• Meaningful founding equity, plus a salary range shared early in the process.
• A founding seat with real ownership: the agentic layer is yours end to end.
• A live design partner, not a hypothetical market: your work ships against a real factory and real people from week one.
• San Francisco Bay Area, hybrid with in-person time as the default. Occasional travel to the design partner is possible, never required.